|Home | About | Journals | Submit | Contact Us | Français|
The structures of peptide collision-induced dissociation (CID) product ions are investigated using ion mobility/ mass spectrometry techniques combined with theoretical methods. The cross-section results are consistent with a mixture of linear and cyclic structures for both b4 and a4 fragment ions. Direct evidence for cyclic structures is essential in rationalizing the appearance of fragments with scrambled (i.e., permutated) primary structures, as the cycle may not open up where it was initially formed. It is demonstrated here that cyclic and linear a4 structures can interconvert freely as a result of collisional activation, implying that isomerization takes place prior to dissociation.
Although it is generally understood that the amino acid chains that comprise peptides in solution are highly dynamic, one does not expect a change in the primary structure (i.e., amino acid sequence) of these molecules to occur. However, in the gas phase, where it is possible to add substantial internal energy, the fragments of these linear structures can undergo rearrangement processes in which the sequence appears to change.1,2 This is especially true when peptide ions are subjected to collisions with a background gas to induce collision-induced dissociation (CID). Generally, the masses of the CID product ions are expected to reflect the amino acid sequence of the peptide and therefore be valuable for sequencing.3,4 In fact, this approach is now a cornerstone in the identification of proteins by mass spectrometry.5,6 The interpretation of these spectra is often carried out using computer algorithms.7,8 With current methods a large fraction of examined ions cannot be identified.9 At least in part, this is due to an incomplete understanding of the underlying fragmentation chemistry.10
As background, it is well-known that under CID conditions, protonated peptide ions generally display abundant cleavages at the backbone amide bonds, giving rise to N-terminal b and C-terminal y product ions.11 In proposed mechanisms, the ionizing proton is thought to be moved from a site of higher proton affinity, such as a basic side chain (e.g., guanidine group of arginine), to the backbone C=O and then NH groups, thereby weakening the amide bond.12,13 Although there has been controversy about the dissociation mechanism and the types of structures that are formed (especially for b fragment ions),14–18 it is now generally accepted that the most likely dissociation pathway occurs via nucleophilic attack from a backbone carbonyl oxygen to a carbonyl carbon, thus forming a five-membered “oxazolone” ring b fragment structure.16,18 This is exemplified in the reaction chemistry of the pentapeptide Leu-enkephalin (Tyr-Gly-Gly-Phe-Leu) that we have chosen in this study (see Scheme 1A, B). Detailed ab inito calculations have confirmed that this pathway is both energetically and entropically accessible.10 Moreover, the presence of oxazolone-type structures has recently been corroborated on the basis of structurally diagnostic oxazolone C=O stretch modes by infrared (IR) spectroscopy.19,20 Linear oxazolone structures are believed to be able to isomerize to fully cyclic structures, following a nucleophilic attack by the N-terminus on the oxazolone ring (Scheme 1C).2 If the cyclic structure opens up at a different amide bond than where it was originally formed, this results in oxazolone structures with permutated (i.e., scrambled) primary structures (gray box, Scheme 1D). The subsequent CID of these scrambled structures is responsible for the appearance of “nondirect” CID fragments,2 which cannot be explained on the basis of the original amino acid sequence of the peptide.
Experimental evidence for such cyclic structures comes from IR spectroscopy; however, the assignment is less definite than in the case of oxazolone structures, and it is difficult to draw conclusions on the relative abundances of the different isomers in the mixture.20 Other indicative evidence for cyclic structures comes from H/D exchange studies21 and ion mobility measurements.22 It should be noted that cyclic configurations are not limited to b ions but can also occur for other N-terminal fragments, such as for instance a fragments (i.e., the b ion less a CO). In analogy to b4, Leu-enkephalin a4 is thought to adopt both linear imine (E) and cyclic structures (F) (Scheme 1).20 A re-opening of the cyclic isomers can again give rise to permutated oxazolone-type structures (gray box G), the subsequent CID of which rationalizes nondirect sequence ions.
Here, we present a systematic study of the structural products that are formed in collision-induced dissociation of the pentapeptide Leu-enkephalin by ion mobility/mass spectrometry.23,24 These products are also compared to the fragments for N-acetyl Leu-enkephalin. We address the question of the relative abundances of the different structural isomers, and whether the relative abundances can be influenced by the amount of energy deposited in the molecule. This result provides insights into the dynamics of the isomerization pathways, which are important in peptide scrambling processes. This is also the first study that uses ion mobility in combination with theoretical approaches to answer structural questions on CID fragment ions.
The protonated pentapeptide Leu-enkephalin (Tyr-Gly-Gly-Phe-Leu) (Sigma Aldrich, St. Louis) precursor ion was generated by electrospray ionization (ESI).25 The ions were trapped in an rf ion funnel prior to pulsed extraction into an intermediate pressure He-filled drift tube (~3 Torr). Ions travel the length of the drift tube under the influence of an electric field (~9 V cm−1) but are retarded in their motion due to collisions with the He background gas. Ions of different collision cross-sections are thus separated in space and will arrive at the detector at different times, defined as the drift time or arrival time distributions.
The laboratory-constructed drift tube design employed in these experiments has been described in detail previously.26 Briefly, it is made up of three consecutive drift tubes, each 0.9 m in length, and the series is followed by an orthogonal time-of-flight mass analyer. At the end of each drift tube, the radially diffuse ion population is re-focused using an rf ion funnel, which is operated at the same pressure as the drift tube. This improves the overall transmission of ions through the drift tube.27 Moreover, by selecting ions of a specific mobility at the end of the first drift tube it is possible to select specific conformers, re-focus these selected ions in the ion funnel, and collisionally activate the mobility-selected ions for injection into the second drift tube.28,29 Additional mobility selections and activations can be carried out in each consecutive drift tube, thus allowing a multistage IMS approach,28 in analogy to multiple mass isolations (MS/MS).
Singly protonated Leu-en-kephalin (m/z 556) was determined to have an experimentally derived collision cross-section Ω ~162 Å2, based on the drift time in the first drift tube. The procedure to calculate cross-sections has been described in detail many times before23,24 and is based on the relation
where z, e, kB, mI and mB, and N denote fundamental properties such as the charge state, elemental charge, Boltzmann’s constant, mass of the ion and mass of the buffer gas, and the neutral number density, respectively. The experimental parameters E, L, P, T and tD correspond to the applied electric field, length of the drift tube (m), pressure (Torr), temperature (K) and arrival time (s).
Under the experimental conditions used, the precursor Leu-enkephalin ion was efficiently mobility-selected by using a 100 μs-wide voltage gating window at the end of the first drift tube (after ~11.8 ms drift time). This precursor ion was then activated by applying a voltage drop between two adjacent rings (up to 230 V over a 3 mm spacing) at the end of the ion funnel,26 thus generating a number of CID product ions, of which y4 (m/z 393), a4 (m/z 397) and b4 (m/z 425) were among the most abundant. The experimental collision cross-sections of the Leu-enkephalin a4 and b4 ions were calculated from the drift time distributions in the second drift tube. This drift time could be accurately determined on the basis of the mobility selection of the precursor Leu-enkephalin at the end of the first drift tube, the known mobility of the precursor up to the activation region at the end of the ion funnel and the mobility selections of the CID fragments at the end of the second drift tube. The “unknown quantity” in this approximation is the temporary change in mobility due to collisional activation and the unknown time for complete thermalization. Nonetheless, the decrease in the drift time of the precursor ion due to collisional activation could be accurately determined and this serves as a good estimate for the “unknown quantity” above. It was found that this decrease in drift time was on the order of 100–200 μs for drift times of ~10 ms through 1 drift tube. This decrease in drift time was taken into account when calculating the cross-sections, thus bringing the total error of the measurement in the range of 1–2%, as commonly expected in ion mobility measurements.30,31 The CID product ion drift time distributions of Leu-enkephalin were also compared to those for N-acetylated Leu-enkephalin. The latter peptide was made by using acetic anhydride to couple an acetyl group to the N-terminus.
A detailed theoretical investigation of the fragment structures for b4 and a4 of Leu-enkephalin has been published before,20 and the density-functional theory geometries (B3LYP/6-31+G(d,p)) and atomic charges of the optimized structures were used to calculate theoretical cross-sections. We used the trajectory method in the MOBCAL program developed by the Jarrold group, which includes He-ion long-range interactions.32 In the nomenclature of the a4 imine structures, the same notation was used as in the previous publication.20 “IM TR N-term-prot” refers the trans imine structure that is protonated on the N-terminus. “IM CI imine-prot” denotes the cis imine structure that is protonated on the imine nitrogen. “Cyc N-prot” refers to the fully cyclic structure of a4, which is protonated on the nitrogen that was previously included in the imine group. The chemical structures are also displayed in Scheme 1.
Protonated Leu-enkephalin (m/z 556) and N-acetyl Leu-enkephalin (m/z 598) were mobility-selected in the same experiment and subjected to collision-induced dissociation under identical conditions, giving rise to their respective b4 (m/z 425 and 467) and a4 (m/z 397 and 439) fragments. The total drift time distributions of each of these fragments are presented in Figure 1; note that the total drift time corresponds to the sum of the drift times that the ions spend in different parts of the split drift tube design. The longer drift times for the N-acetylated Leu-enkephalin fragments, relative to the Leu-enkephalin fragments, are not surprising, given the bulky acetyl group. However, the differences between the b and a arrival time distributions are much more telling. The b4 fragments of Leu-enkephalin (Figure 1C) and N-acetyl Leu-enkephalin (Figure 1D) display relatively narrow peaks, whereas the distributions for a4 are much broader, and clearly involve multiple structures. The drift time distribution of other CID products, such as y4 (Figure S2), also only show one main feature. It is interesting to note that the drift time distribution of Leu-enkephalin a4 in Figure 1 has at least three identifiable peaks (centered at ~31.6, 32.2 and 33.45 ms) but N-acetyl Leu-enkephalin a4 just exhibits two main features (centered at 34.6 and 35 ms). Admittedly, the third peak at 33.45 ms is of low intensity, but it was reproduced independently on multiple measurements (see also Figure S2, Supporting Information). Other studies have also shown that multiple CID product ion structures can be resolved by ion mobility,22,28 though no theoretical analysis was done in those studies. Our assignment of the mobility-separated structures in Figure 1 is tentatively indicated, and each of these will be discussed in detail in the following section.
The dissociation chemistry of Leu-enkephalin and N-acetylated Leu-enkephalin is expected to be quite different: due to the presence of the N-acetyl group, fragments of N-acetylated Leu-enkephalin cannot adopt a fully cyclic structure (see Scheme S1, Supporting Information). Moreover, proton attachment on the N-terminus is no longer possible, thus making the imine N unambiguously the most basic site in the a-ion structure. This limits the number of putative structures that have to be considered to the cis and trans imine configurations. It is thus tempting to assign the two mobility-separated peaks as cis and trans imine, as we have done in Figure 1B. It is worth noting, however, that the trans form is energetically more favored than the cis one.20 The above assignment clearly contradicts this energetics, because it suggests that the cis form is more favored than the trans structure. Further theoretical investigations are underway in our laboratories to resolve this controversy. Preliminary molecular mechanics modeling suggest that there is no significant difference in the predicted collision cross-sections between cis and trans imine structures (not shown).
Figure 2B shows the mobility distribution of Leu-enkephalin a4 on a collision cross-section scale. This mobility distribution was derived from the experimental drift time distribution of a4 (Figure S3) and mobility selections (1–3) at the end of the second drift tube. The corresponding cross-section for each selection was determined from the drift time through the second drift tube (as described in the Methods), thus allowing a conversion of the total drift time distribution into a cross-section scale. The mobility selections are displayed in Figure 2B and the experimental features are labeled as I, II and III for the sake of the discussion hereafter.
For Leu-enkephalin a4 many more structural possibilities have to be considered compared to the N-acetylated variant, including imine structures with proton attachment to the N-terminus, the cyclic conformer, as well as oxazolone-type a4 structures (see Scheme 1 and nomenclature in the Methods). A detailed study of many of these conformers has already been presented,20 and we make use of the lowest-energy structures from that study to try to explain the results shown here. The theoretical cross-sections of the four lowest-energy conformers of each imine variant are presented in Figure 2A, as well as the six lowest-energy conformers of the cyclic structure. The average theoretical collision cross-sections and the standard deviations for each structural variant of a4 are also summarized in Table 1. On the basis of the IR spectroscopy study,20 N-terminal-protonated imine and cyclic structures could be confirmed, whereas imine-protonated and oxazolone-type a4 were excluded. Note that the latter oxazolone-type a4 structures were not excluded on the basis of the calculated structures, but rather on the absence of an oxazolone CO stretch in the IR spectrum for a4.20
A comparison of the calculated cross-sections to the experimental results shows that the most compact structure (I) could be explained by the cyclic geometry (orange pentagons), whereas the most extended conformer (III) matches the trans imine N-terminal-protonated structure (red triangles).26 Feature II could be rationalized by the cis imine N-terminal protonated structures (blue dots). It is nonetheless odd that the higher-energy cis imine should be much more abundant than trans imine, the relative contributions of the different structures being estimated at 31% I, 61% II and 8% III. We are currently performing further calculations to resolve this contradiction.
The theoretical approach followed here involves generating zero-energy structures and corresponding predicted collision cross-sections, which are then compared to experimentally determined cross-sections. This is probably not an accurate reflection of the dynamic structures that are actually probed by ion mobility. As one compares the theoretical mobilities of the lower- to the higher-energy imine structures, one observes a general decrease in the calculated cross-sections (i.e., increase in mobility). This is contrary to what one may expect intuitively, but it suggests that the 300 K mobilites may actually be higher than what is predicted for the lowest-energy zero-K structures. The cyclic configurations, on the other hand, have consistently high mobilities. This is in fact expected from their compact structures and constrained flexibility compared to linear structures. The large range in predicted cross-sections for the linear imine structures is probably due to the combination of a highly flexible backbone around the glycine residues and the presence of bulky aromatic side chains (Phe and Tyr) near the termini, which make an accurate prediction of the collision cross-section challenging. In summary, the structural assignment of the most compact a4 ions as cyclic structures and the most extended structures as linear trans imine structures is relatively convincing; however, the interpretation of feature II remains to be clarified.
The mobility distribution of Leu-enkephalin b4 on a cross-section scale is shown in Figure 3B. The procedure to derive cross-section distributions from drift time distributions is again the same as for a4. The predicted cross-sections and relative energies of five oxazolone and three cyclic structures are displayed in Figure 3A and their average cross-sections are summarized in Table 1. The experimentally derived cross-sections agree well with the predicted cross-sections for the cyclic structure but give an inadequate match to the lowest-energy oxazolone-type structures. This is surprising, as oxazolone structures have been identified by IR spectroscopy for this particular b4 fragment.19,20 Under the CID conditions here, similar fragment structures would be expected to be observed as in that previous study. The higher-energy calculated oxazolone conformers exhibit increased mobility, in analogy to the higher-energy imine a4 conformers, thus at least approaching the experimentally determined cross-sections. In conclusion, it is highly probable that oxazolone structures are formed in the case of b4, which also suggests that the theoretical cross-sections from zero-energy structures consistently overestimate the actual cross-sections of the linear fragments at room temperature.
An argument in favor of a mixture of oxazolone and cyclic structures for b4 can be made in light of the broader peak in the drift time distributions for Leu-enkephalin b4 (Figure 1C) compared to N-acetyl Leu-enkephalin b4 (Figure 1D) under identical experimental conditions. The broader peak shape of Leu-enkephalin b4 is consistent with a mixture of structures, in contrast to N-acetyl Leu-enkephalin b4, where the cyclic structure cannot be formed. It is also interesting that the weak feature at 127 Å2 is consistently reproduced, although a structural assignment could not be made yet.
To assess the influence of the energy deposited in the CID process on the relative abundances of the CID product ions, the activation energy was varied from 180 to 220 V, as shown in Figure 4A, B. At the lower activation energy, the fragmentation yield is low and b4 dominates, whereas at the higher activation energy other products (b3, y3, y4 and a4) become competitive. This has in fact been shown in other studies2,33 and does therefore not come as a surprise. On the other hand, if one compares the drift time distributions of a4 as a function of the activation voltage (see Figure 4C, D), one sees that the distributions are essentially identical! The Boltzmann distributions of the collisionally excited ions are expected to be broad, but as the ions exit the activation region, they are quickly thermalized due to collisions with the background He gas (3 Torr He). As the re-thermalized ion populations display the same distribution of structures, independent of the activation energy for CID, this implies a complete equilibration of structures prior to cooling.
The mobility-selected a4 ions at the end of the second drift tube were subjected to activation (220 V) prior to injection into the third drift tube (activations 1–3; see Figure S4). Although some a4 ions dissociated to smaller CID products, a considerable proportion remained as unfragmented a4 ions. It is then interesting to see what happens to the mobilities of the selected a4 ions as a result of the activation. This is shown for the three selections in Figure 5A1–3.34 Of these selections only selection 3 can be considered a clean isolation, where no other ions are present in the selected ion population. Note that mass isolations are not possible on this instrument, and as some fragment ions have partially overlapping drift times (notably a4, b4 and y4; see Figure S2) this limits their complete isolation. Upon activation, the vast majority (i.e., 90%) of the extended linear imine (isolation 3) is converted to higher-mobility structures, whose cross-sections are compatible with features I and II. In fact, the remaining ratio of III (10%), compared to I and II, after selection and activation is very similar to the ratio of III after CID of the precursor ion (8%). For the two other selections, similar trends are seen, although one needs to be cautious due to the interference from CID of b4. For selection 1, the activation results in the re-appearance of structures with reduced mobility, whose cross-sections match those of feature II. For selection 2, a substantial re-appearance of higher-mobility structures occurs, but also some appearance of lower-mobility ions, which suggests a re-appearance of features I and III. All of these results are compatible with the picture that the individual structures can interconvert freely upon activation. Some readers will notice the similarity of the above approach to spectroscopic methods in supersonic expansions, where particular conformers are selected and activated, causing them to isomerize to other conformers.35
Activation of the mobility-3-selected ions yielded the CID mass spectrum shown in Figure 5B. Some of the subsequent CID products are expected from the linear a4 sequence, whereas others are more ambiguous to interpret (indicated by their masses), as they could be assigned to either direct or nondirect sequence ions.
The observed mobility shifts upon activation of linear a4 can be rationalized by interconversion of the linear, cyclic and unassigned structures. This suggests a complete isomerization and equilibration of the structures prior to dissociation, which is comparable to the conclusions from a previous CID study on mobility-selected ubiquitin conformers.36
To summarize our findings, a semiquantitative potential energy surface for a4 is shown in Figure 6. The barriers to interconversion between different types of structures of a4, linear imine, cyclic and unassigned a4, are lower in energy than dissociation to subsequent fragments. Note that the transition state structures and energies for YGGFL a4 interconversions are not known; however, those for a5 of YAGFL-NH2 have been calculated in detail.2,37 In those calculations the height of the barrier for imine to cyclic conversion was found to be 13.1 kcal mol−1. In Figure 6, it is assumed that the energetics for YGGFL are similar for those of YAGFL-NH2.
The ion mobility results for Leu-enkephalin and its CID fragments corroborate an earlier IR spectroscopy study20 that suggested a mixture of linear and cyclic b4 and a4 structures are formed. Here, the relative contributions of linear and cyclic structures could be approximated for a4, suggesting that 31% adopt a cyclic structure, whereas 8% adopt an extended linear imine structure. The remaining 61% (feature II) have cross-sections that could be rationalized with cis imine structures, although the calculated energies of these structures appear to be too high to account for their predominance in the spectrum. In the case of b4 the calculated cross-sections for the cyclic structures are more consistent with the experimentally determined mobility, even if the cross-sections of higher-energy oxazolone structures did approach those of cyclic structures. As oxazolone structures were confirmed in previous IR spectroscopy measurements, this strongly suggests that zero-energy structures underestimate the mobilities of room-temperature linear ions. In terms of feature II for a4 it is probable that other, yet unidentified, structures account for its large share of the ion population. It is clear from this and other studies that the dissociation chemistry of peptides is highly complex and that many reaction pathways have to be considered, and only some of those were considered here.
These results also underline the importance of validating some of the ion mobility results with other structural techniques, such as isotope labeling38 and IR spectroscopy. In the case of peptide CID fragment ions the chemical information available from IR spectroscopy would be highly complementary and may confirm the structural identity of, e.g., feature II of this study. Multiple gas-phase structural tools are necessary to address the structural complexity of peptide fragment ions, as they can differ in the chemical moieties that are formed and the site where the proton is attached. Moreover, the appearance of mixtures of fragment ions is likely not be limited to Leu-enkephalin, as others have observed multiple structures for particular CID product ions by ion mobility.22,28
Finally, this study showed that the relative abundances of a4 structures formed in CID are independent of the total energy deposited in the molecule, though this clearly does affect the individual fragment ion abundances in the CID mass spectrum. Moreover, extensive isomerization of a4 structures takes place during activation, thereby apparently preventing structure-dependent CID. In other words, the molecules lose all memory of the conformation that they had prior to collisional activation, which could cause scrambling of the primary structure of the peptide to occur independent of the starting structure of the peptide.
We gratefully acknowledge support from the NIH (NIH#R01-AG-024547) and NSF (NSF#CHE-0078737) that allowed for development of instrumentation used in this study. N.P. thanks the University of Florida for providing generous start-up funds. B.P. acknowledges the DFG for their support (SU 244/3-1).